[ggmarshall]

Summary

Extends the LQ calibration to handle multiple LQ parameters, mirroring the multi-parameter A/E schema from #42: run_lq_calibration loops over config["params"] — one entry per LQ parameter, each with its own lq_param (no longer hardcoded lq80), energy_param, and optional dt_param/cdf/suffix/plot_options. Results nest per entry under the lq key, pickled objects as {name: LQCal}, plots per name. Config shape:

run_lq: true
threshold: 800
cal_energy_param: cuspEmax_ctc_cal
cut_field: is_valid_cal
params:
  lq:
    lq_param: lq80
    energy_param: cuspEmax
  lq_alt:
    lq_param: lq80
    energy_param: cuspEmax
    suffix: alt

Suffixed entries require LQCal.calibrate suffix support (legend-exp/pygama#707); the kwarg is only passed when a suffix is configured, so unsuffixed entries keep working with released pygama. The par-geds-hit-lq entry point builds its column list from all entries and now stores just the {name: LQCal} mapping under the pickle's lq key instead of re-nesting the full object dict.

The second commit pre-resolves the aoe.py divergence with main: it brings the params-loop run_aoe_calibration from #42 into this lineage combined with the use_log_pdf knob, and carries the schema fixes from #46.

Stacked on #44 (perf-par-dsp) — only the two commits on top are new.

Test plan

Full suite (unit + skimmed-data integration hit chain, ecal→aoe→qc→lq): 71 passed locally, with the LQ config exercising an unsuffixed primary plus a suffixed duplicate entry end to end, and the aoe config migrated to the params schema. Companion legend-dataflow pht chain (partcal→aoe→lq→fast) also passes.

Per AI_POLICY.md: developed with AI assistance (Claude); reviewed by the submitter.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR extends the GEDS hit-level LQ and A/E calibrations to support the multi-parameter params config schema (multiple calibrations per run), including optional suffixing to avoid output-name collisions and updated result/object nesting.

Changes:

• Update run_lq_calibration and par-geds-hit-lq to iterate over config["params"], supporting per-entry lq_param, energy_param, optional dt_param/cdf/suffix/plot_options, and storing {name: LQCal} under the pickle’s lq key.
• Update run_aoe_calibration and par-geds-hit-aoe to use the same params schema, preserve use_log_pdf, and nest results/objects/plots per params entry.
• Adjust docstrings and CLI-required config keys to match the new schema.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
src/legenddataflowscripts/par/geds/hit/lq.py	Convert LQ calibration to params-driven multi-calibration flow; update CLI inputs and outputs accordingly.
src/legenddataflowscripts/par/geds/hit/aoe.py	Convert A/E calibration to params-driven multi-calibration flow while retaining use_log_pdf; update CLI inputs and outputs accordingly.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ggmarshall]

Agreed — fixed in 197de2a. Added a require_unique_suffixes helper in utils/cfgtools and called it from run_lq_calibration (and run_aoe_calibration) right after the top-level require_config_keys. Two entries resolving to the same suffix — including the two-entries-with-no-suffix case you call out — now fail up front with the duplicated values named (<no suffix> for the None case) rather than silently overwriting LQ_Ecorr/cal-dict keys. Unit-tested in tests/test_utils_helpers.py::test_require_unique_suffixes.

[ggmarshall]

Fixed in 197de2a — same require_unique_suffixes guard is now applied in run_aoe_calibration before any calibration runs, so duplicate (or duplicate-absent) suffixes fail fast instead of silently overwriting AoE_Uncorr/cal-dict entries from earlier iterations. See the sibling comment on lq.py for details.

[ggmarshall]

Good catch — that was a genuine bug, not just a naming oddity: with debug_mode set in a config, compt_bands_width received True/False instead of a width. Fixed in 197de2a:

compt_bands_width=param_config.get(
    "compt_bands_width", config.get("compt_bands_width", 20)
),

per-params-entry with a top-level fallback and the original default of 20. (debug_mode itself is still read separately on the line below, unchanged.)

[ggmarshall]

Already addressed on this branch by a3dd041 — the CLI validates each entry with require_config_keys(param_config, ["current_param", "energy_param"], f"aoe config params entry '{name}' ({args.config_file})") before indexing them into the column list, so schema errors name the entry and config file consistently with run_aoe_calibration.